Abstract 14274: Machine Learning-Enhanced Determinants of Oral Anticoagulation in Atrial Fibrillation: Findings From the PINNACLE Registry

Abstract

Introduction: Current atrial fibrillation (AF) guidelines recommend use of oral anticoagulation (OAC) to reduce the risk of stroke and systemic embolism. Machine learning (ML) techniques may enhance our understanding of factors associated with gaps in real-world prescription of OAC. Hypothesis: There will be large variation in county-level OAC prescription rates among ambulatory patients with AF, and geographic and medication factors will be stronger predictors than clinical comorbidities of lack of OAC prescriptions. Methods: Between January 2017 and June 2018, we identified patients with AF from the American College of Cardiology’s (ACC) Practice Innovation and Clinical Excellence (PINNACLE) Registry. We examined associations between patient and site-of-care factors and prescription of OAC across U.S. counties. Several ML methods were used to identify factors associated with lack of OAC prescription. Results: Among 864,339 patients with AF, 586,560 (68%) were prescribed OAC. County OAC prescription rates ranged from 26.8% (15/56) to 93% (41/44), with higher OAC use in the Western US. Supervised ML analysis (extreme gradient boosting: area under the receiver operating characteristic curve [AUC]: 0.811, 95% confidence interval [CI]: 0.809-0.813) outperformed the CHA2DS2-VASc score (AUC: 0.571, 95% CI: 0.569-0.574) in predicting likelihood of OAC prescriptions and identified a rank order of patient features associated with OAC prescription. In the ML models, use of aspirin, antihypertensives, lipid modifying agents, antiarrhythmic agents, age, inter-normalized ratio, median household income, and clinic size were the most important predictors of OAC prescription. Conclusions: In a national cohort of patients with AF underuse of OAC remains high, with notable geographic variation. ML algorithms can identify factors associated with use of OAC.